Suppose you have built an undersea or other underwater habitat, like Jacques Cousteau, Tom Swift, or many popular science articles from the 1950s; but it's autonomous rather than having an umbilical to a surface support station. How do you get energy to run and repair your base?
Fission power is clearly the best option, but suppose you can't use fission power for political reasons.
If you have sunlight, you can of course use solar panels, although the water will absorb some of it. (See Notes and calculations on building luxury underground arcologies for whoever wants them for notes on the absorption spectra of different kinds of water.) If you're under enough water to make it difficult to see you from the surface, or from space, you will get only a tiny amount of sunlight energy.
You can periodically receive shipments of fuel, but burning fuel requires an oxidizer such as oxygen, so you need shipments of oxidizer too. As described in The Suburbean: a minimally-mobile dwelling machine with months of autonomy, the best option to me seems to be sodium chlorate, but even so, you end up with only 5 MJ/kg when you include the weight of the sodium chlorate rather than the usual 43 MJ/kg we're used to on land.
By floating a buoy on the surface, it's possible to gather surface solar energy, and also some amount of wave power, as well as sucking in air for human respiration and perhaps fuel combustion. But the buoy is visible, which may be undesirable, and also exposed to damage from heavy seas, ship collisions, oxygen, and sunlight.
I think the best option in many places is analogous to kite wind power, using a reconfigurable-geometry buoy floated some distance below the surface but above the bed; when configured for high-drag geometry, it pulls on its tether, which is gradually let out, generating power; when the tether is nearly exhausted or it's too close to the surface, it's reconfigured for a lower-drag geometry and reeled back in at a much lower energy cost than what it generated when it was being let out. This is much easier than the equivalent task in the air because the velocities are lower, the forces are much higher, the rope length and therefore snapback potential energy is lower, and it's easy to reconfigure the "kite's" buoyancy for a particular altitude under the water.
(I think passive altitude control via buoyancy control is much more difficult in water than in air; the density of air varies sufficiently with pressure for a balloon to hover within tens of centimeters of a constant height as the rubber holds its own volume relatively constant; on the other hand, in a submarine, water density varies only very slightly with pressure, but higher pressure will tend to collapse your swim bladder and reduce your buoyancy further.)
Cables under tension can carry an amazing amount of power. Consider gel-spun UHMWPE, with its 2.4 GPa yield stress (see Dyneema). At 10 m/s, a snappy but not insane speed (22 miles per hour, in medieval units), 2.4 GPa is 24 GW/m², which is 24 kW/mm². According to Induction kiln, AWG20 [copper] wire can safely carry 5 amps and is 0.812 mm in diameter (not counting the insulation), or 9.7 megawatts/volt/m²; so, reaching the same 24 GW/m² with it requires 2500 volts. At 100 m/s, the UHMWPE cable carries 240 GW/m² = 240 kW/mm², which requires 25 kV in the electrical wire. Copper weighs 9 g/cc, about 9× what UHMWPE weighs.
Lower cable speeds require proportionally more tension and thus more cable thickness to deliver the same power.
A regular air kite might be better in some sense, particularly at those underwater sites, such as those at the bottom of medium-sized lakes, that have no significant water currents. It could be made of a hydrophobic material, floated to the surface of the water by slight inflation, and then floated to kite height by further inflation with hydrogen or helium. Once in the air it can expand substantially to a size much larger than the underwater structure.
In The Suburbean: a minimally-mobile dwelling machine with months of autonomy, it is proposed to store some 500 MJ of energy in half a tonne of Li-ion batteries to provide a month's worth of 180-watt autonomy without access to air, in a habitat equipped with, among other things, 25 kW of winches. Suppose that the kite pulls 100 kilonewtons, which is a bit over ten tonnes (42 mm² of UHMWPE, or maybe 100 mm² = 1 cm² to have a safety factor), and rises to a height of only 200 m in order to avoid interfering with aircraft; and suppose that the wind at that height is generally 20 m/s. That provides a megawatt or two of power, enough to fully charge the batteries in five to twenty minutes, although it requires some kind of apparatus in the lake habitat capable of storing megawatts of power.
So, like a ball python, this underwater habitat could lurk unobserved at the bottom of a still, dark lake, reaching out to replenish its energy from its environment once or twice a month; but instead of swallowing a rat, it silently flies a kite for a few minutes in the middle of the night.
As I described previously in "hovering kite assemblages" (?) a flying machine large enough to simultaneously be at altitudes with winds in different directions can use those differing wind directions to maintain tautness in the tethers between its various otherwise disconnected parts, to control its direction of movement, and to generate energy to power it, in particular keeping it from falling out of the sky. Similarly, a group of tethered submarines at different depths could harness the differing directions of deep-sea currents at those depths to generate energy and control their direction of movement. (Lift in that case is unnecessary.)
In a sense, that's what a sailboat is doing: harnessing energy from the relative movement between the air and water to move in any direction, including tacking upwind.
The hydrodynamic force of a fluid flowing past a body can act on it in almost any direction; in general it is proportional to the density of the fluid, the square of the impact velocity, and the cross-sectional area of the body perpendicular to the direction of the flow, and has an additional factor which I think is actually the "drag coefficient". Conventionally it's resolved into a scalar in the direction of the fluid flow called "drag" and a two-dimensional vector perpendicular to it called "lift".
In a constant flow in the absence of any other applied force, drag gradually accelerates the body to the velocity of the flow, causing the hydrodynamic force to disappear. This leads to an interesting phenomenon where the power produced at zero velocity and maximum force is zero, and the power produced at zero force and maximum velocity is also zero; maximum power is somewhere in between, specifically at one third of maximum velocity. (Unless the drag coefficient changes, which it does.)
Because drag is complicated --- the coefficients vary with flow speed and viscosity, and not even continuously or monotonically --- I hesitate to pronounce anything too pompous about this, but very roughly this suggests that water currents tend to produce about a thousand times as much force and power as wind of the same speed; for wind to provide the same force, it needs to go 32 times as fast, but to provide the same power, it only needs to go 10 times as fast. Or, changing a different variable, wind needs a thousand times as much area to press on to be equivalent to a water current of the same speed.